Staff Flight Research Machine Learning Engineer

Joby Aviation•Santa Cruz, CA

10d•$224,000 - $245,000

About The Position

Joby Flight Research designs, develops, and flight-tests novel aircraft using a software-first autonomy approach. We build and deploy autonomy, perception, planning, and radar systems across conventional, electric, and hydrogen-electric aircraft in both CTOL and VTOL configurations. Joby Flight Research is seeking a Staff Machine Learning Engineer to design and build state-of-the-art perception and reasoning algorithms into the Superpilot™ autonomy stack, enabling autonomous aircraft and associated ground systems to safely and autonomously navigate their complex environment, and define the standard of autonomous flight. In this role, you will be responsible for training models that enhance current Superpilot™ algorithms while investigating the application of cutting-edge research to expand autonomous flight capabilities. Your responsibilities will include filtering sensor data from flights, architecting dependable training infrastructure for datasets and models, and performing ongoing evaluations within our Operational Design Domain (ODD). By focusing on these tasks, you will play a vital part in achieving the detection and localization standards necessary for ensuring the safety of autonomous aviation. We are a small, high-impact team that values curiosity, technical initiative, and the ability to operate independently. You will collaborate deeply with controls, and flight software engineers to build a foundation that accelerates our path to safe, autonomous flight. The right candidate is a strong ML engineer who cares deeply about system integration, experiment reproducibility and traceability, and is able to contribute to different parts of the data pipeline where needed.

Requirements

At least 8 years of experience developing and implementing advanced perception frameworks for autonomous platforms, including aircraft, vehicles, or robotics
Extensive practical knowledge of cutting-edge models for tracking and detecting objects within real-world applications
Comprehensive understanding of geometric vision methods such as visual odometry, structure-from-motion, and stereo vision to facilitate accurate 3D estimation
Competency in C++, Python, and PyTorch, along with various deep learning inference and training ecosystems
Ability to rapidly prototype code during experimentation, and support deployment by shipping scalable and high quality code
A creative approach to engineering that includes a history of mitigating performance delays and addressing complex operational edge cases
Direct experience maintaining commercial autonomous systems by diagnosing technical hurdles and improving live system dependability
Strong interpersonal skills necessary to succeed within a fast-paced, multidisciplinary team focused on safety-critical engineering

Nice To Haves

Expertise in developing supporting data curation and model experimentation pipelines to support ML experimentation
Experience with ROS 2 or other robotics middlewares
Experience with recent generative AI world simulation tooling (e.g. NVIDIA Isaac Sim, Omniverse)
Experience deploying models onto GPUs and other accelerators for production-level embedded systems
Experience with autonomous vehicles
Experience in processing aircraft data (GPS, inertial, air data, radio data, etc.)
Expert-level software engineering: deep expertise in architecting and writing clean, scalable, and maintainable code
Experience with version control and CI/CD platforms, able to manage your software through its entire lifecycle (development, testing, deployment)
Experience deploying ML models in a production environment using modern MLOps principles and tools (e.g., MLflow, Kubeflow)
Experience mentoring small teams of perception engineers up to 5 people

Responsibilities

Own the development of a Superpilot™ application (e.g. ground hazard avoidance, vision-based landing, detect and avoid)
Lead product requirements and design architecture with multi-disciplinary teams—including controls, systems, and flight testing—to seamlessly embed and validate algorithms within our operational design domain
Architect and deploy sophisticated algorithms for aircraft environment detection and tracking. By integrating deep learning with geometric computer vision, you will utilize multi-sensor inputs—including lidar, radar, and varied camera systems—to establish comprehensive 3D situational awareness
Construct high-performance model training pipelines and extensive evaluation systems. These frameworks must ensure reliability by identifying performance nuances in complex edge cases and rare operational scenarios
Drive continuous improvement of perception stacks through simulation. You will focus on optimizing latency and robustness to ensure peak performance during demanding flight conditions
Engineer specialized diagnostic and visualization tools to extract meaningful insights from field data. This involves facilitating swift root-cause analysis and resolving perception challenges in active deployments
Shape the future of the Superpilot™ autonomy system. Your contributions will define a perception architecture that sets new benchmarks for safety and reliability in the autonomous aviation industry